error in GermlineCombineChannel

[anoronh4]

I found a persistent error (failed 4x) in the process GermlineCombineChannel which seems to be caused by one of the vcf inputs. The .command.err was as follows:

$ cat work/58/3d153601e0bc08cab9e5b0aa43d24e/.command.err
[W::vcf_parse_format] FORMAT '' is not defined in the header, assuming Type=String
[W::vcf_parse_format] FORMAT '' is not defined in the header, assuming Type=String
[W::vcf_parse_format] FORMAT '' is not defined in the header, assuming Type=String
Encountered error, cannot proceed. Please check the error output above.

I used gatk's ValidateVariants and vcftools' vcf-validator, and both found fault with the strelka input, but for different reasons. With gatk, the report said:

A USER ERROR has occurred: Input /juno/work/tempo/PoN/work/05/3bf9a12217be8b3b537fe7d9d6b2c6/<maskedId>.strelka2.vcf.gz fails strict validation of type ALL: one or more of the ALT allele(s) for the record at position 1:12837358 are not observed at all in the sample genotypes

and the variant record shows as follows:

1	12837358	.	G	A	.	NoPassedVariantGTs	SNVHPOL=2;MQ=56	GT:GQ:GQX:DP:DPF:AD:ADF:ADR:SB:FT:PL	0/0:192:192:309:1:246,63:126,48:120,15:0.0:PASS:0,190,370

vcf-tools reports:

Wrong number of fieldsin /juno/work/tempo/PoN/work/05/3bf9a12217be8b3b537fe7d9d6b2c6/<maskedID>.strelka2.vcf.gz; expected 10, got 12. The offending line was:
[7	151927026	.	T	C	1064	PASS	SNVHPOL=3;MQ=33	GT:GQ:GQX:DP:LowGQX;NoPassedVariantGTs	CIGAR=1M6I;RU=GCG;REFREP=9;IDREP=11;MQ=60	GT:GQ:GQX:DPI:AD:ADF:ADR:FT:PL	0/1:11:11:26:2,8:1,3:1,5:LowGQX:139,0,8]

My guess is that the latter is causing the issue, because that seems to be a formatting error. this is not the first issue i've seen related to strelka (#887) and it might be a good idea to run a check like one of the above before exiting the strelka process before issues arise downstream.

[anoronh4]

Fixing this to include a vcf validation step would require changing the current images for SomaticRunStrelka2 and GermlineRunStrelka2. Neither have vcftools installed, so that will need to be added to the images. the command is:

vcf-validator <inputvcf>

no vcftools antecedent required. If there is something wrong with the vcf, a detailed message and a non-zero exit code follow, forcing the Strelka step to repeat itself automatically to produce an vcf that downstream steps can use.

[gongyixiao]

Agree with adding vcftools and using vcf-validator.

Hopefully it can solve #887 as well

[anoronh4]

we need to check if vcf-validator is already in the container